void
DocWriter_add_segment(DocWriter *self, SegReader *reader,
I32Array *doc_map) {
DocWriterIVARS *const ivars = DocWriter_IVARS(self);
int32_t doc_max = SegReader_Doc_Max(reader);
if (doc_max == 0) {
// Bail if the supplied segment is empty.
return;
}
else {
OutStream *const dat_out = S_lazy_init(self);
OutStream *const ix_out = ivars->ix_out;
ByteBuf *const buffer = BB_new(0);
DefaultDocReader *const doc_reader
= (DefaultDocReader*)CERTIFY(
SegReader_Obtain(reader, VTable_Get_Name(DOCREADER)),
DEFAULTDOCREADER);
for (int32_t i = 1, max = SegReader_Doc_Max(reader); i <= max; i++) {
if (I32Arr_Get(doc_map, i)) {
int64_t start = OutStream_Tell(dat_out);
// Copy record over.
DefDocReader_Read_Record(doc_reader, buffer, i);
char *buf = BB_Get_Buf(buffer);
size_t size = BB_Get_Size(buffer);
OutStream_Write_Bytes(dat_out, buf, size);
// Write file pointer.
OutStream_Write_I64(ix_out, start);
}
}
DECREF(buffer);
}
}
static void
S_init_arena(MemoryPool *self, MemoryPoolIVARS *ivars, size_t amount) {
UNUSED_VAR(self);
ByteBuf *bb;
// Indicate which arena we're using at present.
ivars->tick++;
if (ivars->tick < (int32_t)Vec_Get_Size(ivars->arenas)) {
// In recycle mode, use previously acquired memory.
bb = (ByteBuf*)Vec_Fetch(ivars->arenas, (size_t)ivars->tick);
if (amount >= BB_Get_Size(bb)) {
BB_Grow(bb, amount);
BB_Set_Size(bb, amount);
}
}
else {
// In add mode, get more mem from system.
size_t buf_size = (amount + 1) > ivars->arena_size
? (amount + 1)
: ivars->arena_size;
char *ptr = (char*)MALLOCATE(buf_size);
bb = BB_new_steal_bytes(ptr, buf_size - 1, buf_size);
Vec_Push(ivars->arenas, (Obj*)bb);
}
// Recalculate consumption to take into account blocked off space.
ivars->consumed = 0;
for (int32_t i = 0; i < ivars->tick; i++) {
ByteBuf *bb = (ByteBuf*)Vec_Fetch(ivars->arenas, (size_t)i);
ivars->consumed += BB_Get_Size(bb);
}
ivars->buf = BB_Get_Buf(bb);
ivars->limit = ivars->buf + BB_Get_Size(bb);
}
static Hash*
S_extract_tv_cache(ByteBuf *field_buf) {
Hash *tv_cache = Hash_new(0);
char *tv_string = BB_Get_Buf(field_buf);
int32_t num_terms = NumUtil_decode_c32(&tv_string);
CharBuf *text = CB_new(0);
// Read the number of highlightable terms in the field.
for (int32_t i = 0; i < num_terms; i++) {
size_t overlap = NumUtil_decode_c32(&tv_string);
size_t len = NumUtil_decode_c32(&tv_string);
// Decompress the term text.
CB_Set_Size(text, overlap);
CB_Cat_Trusted_Str(text, tv_string, len);
tv_string += len;
// Get positions & offsets string.
char *bookmark_ptr = tv_string;
int32_t num_positions = NumUtil_decode_c32(&tv_string);
while (num_positions--) {
// Leave nums compressed to save a little mem.
NumUtil_skip_cint(&tv_string);
NumUtil_skip_cint(&tv_string);
NumUtil_skip_cint(&tv_string);
}
len = tv_string - bookmark_ptr;
// Store the $text => $posdata pair in the output hash.
Hash_Store(tv_cache, (Obj*)text,
(Obj*)BB_new_bytes(bookmark_ptr, len));
}
DECREF(text);
return tv_cache;
}
void
HLWriter_add_segment(HighlightWriter *self, SegReader *reader,
I32Array *doc_map)
{
int32_t doc_max = SegReader_Doc_Max(reader);
if (doc_max == 0) {
// Bail if the supplied segment is empty.
return;
}
else {
DefaultHighlightReader *hl_reader = (DefaultHighlightReader*)
CERTIFY(SegReader_Obtain(reader,
VTable_Get_Name(HIGHLIGHTREADER)), DEFAULTHIGHLIGHTREADER);
OutStream *dat_out = S_lazy_init(self);
OutStream *ix_out = self->ix_out;
int32_t orig;
ByteBuf *bb = BB_new(0);
for (orig = 1; orig <= doc_max; orig++) {
// Skip deleted docs.
if (doc_map && !I32Arr_Get(doc_map, orig))
continue;
// Write file pointer.
OutStream_Write_I64( ix_out, OutStream_Tell(dat_out) );
// Copy the raw record.
DefHLReader_Read_Record(hl_reader, orig, bb);
OutStream_Write_Bytes(dat_out, BB_Get_Buf(bb), BB_Get_Size(bb));
BB_Set_Size(bb, 0);
}
DECREF(bb);
}
}
ByteBuf*
HLWriter_TV_Buf_IMP(HighlightWriter *self, Inversion *inversion) {
const char *last_text = "";
size_t last_len = 0;
ByteBuf *tv_buf = BB_new(20 + Inversion_Get_Size(inversion) * 8);
uint32_t num_postings = 0;
Token **tokens;
uint32_t freq;
UNUSED_VAR(self);
// Leave space for a c32 indicating the number of postings.
BB_Set_Size(tv_buf, C32_MAX_BYTES);
Inversion_Reset(inversion);
while ((tokens = Inversion_Next_Cluster(inversion, &freq)) != NULL) {
Token *token = *tokens;
char *const token_text = Token_Get_Text(token);
const int32_t token_len = Token_Get_Len(token);
int32_t overlap = StrHelp_overlap(last_text, token_text,
last_len, token_len);
char *ptr;
char *orig;
size_t old_size = BB_Get_Size(tv_buf);
size_t new_size = old_size
+ C32_MAX_BYTES // overlap
+ C32_MAX_BYTES // length of string diff
+ (token_len - overlap) // diff char data
+ C32_MAX_BYTES // num prox
+ (C32_MAX_BYTES * freq * 3); // pos data
// Allocate for worst-case scenario.
ptr = BB_Grow(tv_buf, new_size);
orig = ptr;
ptr += old_size;
// Track number of postings.
num_postings += 1;
// Append the string diff to the tv_buf.
NumUtil_encode_c32(overlap, &ptr);
NumUtil_encode_c32((token_len - overlap), &ptr);
memcpy(ptr, (token_text + overlap), (token_len - overlap));
ptr += token_len - overlap;
// Save text and text_len for comparison next loop.
last_text = token_text;
last_len = token_len;
// Append the number of positions for this term.
NumUtil_encode_c32(freq, &ptr);
do {
// Add position, start_offset, and end_offset to tv_buf.
NumUtil_encode_c32(Token_Get_Pos(token), &ptr);
NumUtil_encode_c32(Token_Get_Start_Offset(token), &ptr);
NumUtil_encode_c32(Token_Get_End_Offset(token), &ptr);
} while (--freq && (token = *++tokens));
// Set new byte length.
BB_Set_Size(tv_buf, ptr - orig);
}
// Go back and start the term vector string with the posting count.
char *dest = BB_Get_Buf(tv_buf);
NumUtil_encode_padded_c32(num_postings, &dest);
return tv_buf;
}
void
DocWriter_Add_Inverted_Doc_IMP(DocWriter *self, Inverter *inverter,
int32_t doc_id) {
DocWriterIVARS *const ivars = DocWriter_IVARS(self);
OutStream *dat_out = S_lazy_init(self);
OutStream *ix_out = ivars->ix_out;
uint32_t num_stored = 0;
int64_t start = OutStream_Tell(dat_out);
int64_t expected = OutStream_Tell(ix_out) / 8;
// Verify doc id.
if (doc_id != expected) {
THROW(ERR, "Expected doc id %i64 but got %i32", expected, doc_id);
}
// Write the number of stored fields.
Inverter_Iterate(inverter);
while (Inverter_Next(inverter)) {
FieldType *type = Inverter_Get_Type(inverter);
if (FType_Stored(type)) { num_stored++; }
}
OutStream_Write_C32(dat_out, num_stored);
Inverter_Iterate(inverter);
while (Inverter_Next(inverter)) {
// Only store fields marked as "stored".
FieldType *type = Inverter_Get_Type(inverter);
if (FType_Stored(type)) {
String *field = Inverter_Get_Field_Name(inverter);
Obj *value = Inverter_Get_Value(inverter);
Freezer_serialize_string(field, dat_out);
switch (FType_Primitive_ID(type) & FType_PRIMITIVE_ID_MASK) {
case FType_TEXT: {
const char *buf = Str_Get_Ptr8((String*)value);
size_t size = Str_Get_Size((String*)value);
OutStream_Write_C32(dat_out, size);
OutStream_Write_Bytes(dat_out, buf, size);
break;
}
case FType_BLOB: {
char *buf = BB_Get_Buf((ByteBuf*)value);
size_t size = BB_Get_Size((ByteBuf*)value);
OutStream_Write_C32(dat_out, size);
OutStream_Write_Bytes(dat_out, buf, size);
break;
}
case FType_INT32: {
int32_t val = Int32_Get_Value((Integer32*)value);
OutStream_Write_C32(dat_out, val);
break;
}
case FType_INT64: {
int64_t val = Int64_Get_Value((Integer64*)value);
OutStream_Write_C64(dat_out, val);
break;
}
case FType_FLOAT32: {
float val = Float32_Get_Value((Float32*)value);
OutStream_Write_F32(dat_out, val);
break;
}
case FType_FLOAT64: {
double val = Float64_Get_Value((Float64*)value);
OutStream_Write_F64(dat_out, val);
break;
}
default:
THROW(ERR, "Unrecognized type: %o", type);
}
}
}
// Write file pointer.
OutStream_Write_I64(ix_out, start);
}
